Benchmarks

This section describes a range of performance benchmarks that have been run comparing this library with the standard library, and how to run your own benchmarks if required.

The values in the ratio column are how many times longer running a specific operation takes in comparison to the same operation with a double.

On nearly all platforms, there is hardware support for binary floating point math, so we are comparing hardware to software runtimes; Decimal will be slower

Both the results from Intel and GCC types are from very close, but not identical benchmark routines since they are written in C instead of C++. We assume they are close enough, and the differences between the C and C++ compilers are small enough, for fair comparison

How to run the Benchmarks

To run the benchmarks yourself, navigate to the test folder and define BOOST_DECIMAL_RUN_BENCHMARKS when running the tests. An example on Linux with b2: ../../../b2 cxxstd=20 toolset=gcc-13 define=BOOST_DECIMAL_RUN_BENCHMARKS benchmarks -a release, or ../../../b2 cxxstd=20 toolset=gcc-13 define=BOOST_DECIMAL_RUN_BENCHMARKS=1,BOOST_DECIMAL_BENCHMARK_CHARCONV=1 benchmarks -a release to also run the <charconv> benchmarks

To run the GCC benchmarks you can use the following command: gcc benchmark_libdfp.c -O3 -std=c17 followed by: ./a.out

To run the Intel benchmarks you will need both the Intel Compiler, and the library. You can the use the following command: icx benchmark_libbid.c -O3 $PATH_TO_LIBBID/libbid.a -std=c17 followed by: ./a.out

The Intel benchmarks can only be run on one of their supported architectures: IA-32, IA-64, and Intel x64

Methodology

Comparisons

The benchmark for comparisons generates a random vector containing 20,000,000 elements and does operations >, >=, <, <=, ==, and != between vec[i] and vec[i + 1]. This is repeated 5 times to generate stable results.

Basic Mathematical Operations

The benchmark for these operations generates a random vector containing 20 million elements and does operations +, -, *, / between vec[i] and vec[i + 1]. This is repeated five times to generate stable results.

`<charconv>`

Parsing and serializing number exactly is one of the key features of decimal floating point types, so we must compare the performance of <charconv>. For all the following the results compare we compare against STL provided <charconv> for 20 million conversions. Since <charconv> is fully implemented in software for each type the performance gap between built-in float and double vs decimal32_t and decimal64_t is significantly smaller (or the decimal performance is better) than the hardware vs. software performance gap seen above for basic operations.

To run these benchmarks yourself, you will need a compiler with complete implementation of <charconv> and to run the benchmarks under C++17 or higher. At the time of writing, this is limited to:

GCC 11 or newer
MSVC 19.24 or newer

These benchmarks are automatically disabled if your compiler does not provide feature complete <charconv> or if the language standard is set to C++14.

x64 Linux

Run using an Intel i9-11900k chipset running Ubuntu 24.04 and Intel oneAPI compiler 2025.2.0 or GCC 13.3.0.

Comparisons

Intel Compiler:

Type	Runtime (us)	Ratio to `double`
`float`	72,696	0.505
`double`	143,924	1.000
`decimal32_t`	1,485,786	10.323
`decimal64_t`	1,653,991	11.492
`decimal128_t`	4,662,704	32.397
`decimal_fast32_t`	619,662	4.305
`decimal_fast64_t`	606,382	4.213
`decimal_fast128_t`	698,945	4.856
Intel `Decimal32`	2,411,294	16.754
Intel `Decimal64`	3,158,422	21.945
Intel `Decimal128`	3,389,883	23.553

GCC:

Type	Runtime (us)	Ratio to `double`
`float`	56,457	0.916
`double`	61,615	1.000
`decimal32_t`	1,404,638	22.797
`decimal64_t`	1,408,074	22.853
`decimal128_t`	4,974,170	80.730
`decimal_fast32_t`	546,836	8.875
`decimal_fast64_t`	472,387	7.667
`decimal_fast128_t`	480,853	7.804
GCC `_Decimal32`	816,703	13.255
GCC `_Decimal64`	501,479	8.139
GCC `_Decimal128`	914,600	14.844

Addition

Intel Compiler:

Type	Runtime (us)	Ratio to `double`
`float`	118,040	1.303
`double`	90,579	1.000
`decimal32_t`	1,712,396	18.905
`decimal64_t`	1,575,893	17.398
`decimal128_t`	3,181,562	35.125
`decimal_fast32_t`	729,257	8.051
`decimal_fast64_t`	1,083,923	11.967
`decimal_fast128_t`	1,367,004	15.092
Intel `Decimal32`	1,242,797	13.721
Intel `Decimal64`	1,689,585	18.653
Intel `Decimal128`	1,958,345	21.620

GCC:

Type	Runtime (us)	Ratio to `double`
`float`	79,256	1.085
`double`	73,017	1.000
`decimal32_t`	1,501,645	20.566
`decimal64_t`	1,567,250	21.464
`decimal128_t`	4,609,413	63.128
`decimal_fast32_t`	735,864	10.078
`decimal_fast64_t`	1,002,119	13.724
`decimal_fast128_t`	1,329,644	18.210
GCC `_Decimal32`	2,975,146	40.746
GCC `_Decimal64`	2,186,565	29.946
GCC `_Decimal128`	3,368,864	46.138

Subtraction

Intel Compiler:

Type	Runtime (us)	Ratio to `double`
`float`	78,250	1.069
`double`	73,193	1.000
`decimal32_t`	1,480,678	20.229
`decimal64_t`	1,371,677	18.741
`decimal128_t`	2,768,955	37.831
`decimal_fast32_t`	1,040,587	14.217
`decimal_fast64_t`	1,055,980	14.427
`decimal_fast128_t`	1,212,405	16.564
Intel `Decimal32`	1,922,108	26.261
Intel `Decimal64`	1,793,879	24.509
Intel `Decimal128`	2,397,372	32.754

GCC:

Type	Runtime (us)	Ratio to `double`
`float`	275,230	0.936
`double`	293,907	1.000
`decimal32_t`	1,451,610	4.939
`decimal64_t`	1,456,587	4.956
`decimal128_t`	4,332,644	14.742
`decimal_fast32_t`	842,910	2.868
`decimal_fast64_t`	968,939	3.297
`decimal_fast128_t`	1,327,411	4.516
GCC `_Decimal32`	2,045,306	6.959
GCC `_Decimal64`	1,355,777	4.613
GCC `_Decimal128`	3,178,891	10.816

Multiplication

Intel Compiler:

Type	Runtime (us)	Ratio to `double`
`float`	78,445	1.078
`double`	72,798	1.000
`decimal32_t`	1,735,239	23.836
`decimal64_t`	2,272,739	31.220
`decimal128_t`	6,396,750	87.870
`decimal_fast32_t`	993,256	13.644
`decimal_fast64_t`	1,670,141	22.942
`decimal_fast128_t`	5,959,977	81.870
Intel `Decimal32`	1,375,434	18.894
Intel `Decimal64`	2,052,278	28.191
Intel `Decimal128`	5,964,489	81.932

GCC:

Type	Runtime (us)	Ratio to `double`
`float`	76,238	1.161
`double`	65,652	1.000
`decimal32_t`	1,703,365	25.945
`decimal64_t`	2,564,605	39.063
`decimal128_t`	7,115,514	108.382
`decimal_fast32_t`	1,225,047	18.660
`decimal_fast64_t`	1,904,509	29.009
`decimal_fast128_t`	6,056,348	92.249
GCC `_Decimal32`	2,635,531	40.144
GCC `_Decimal64`	2,545,441	38.772
GCC `_Decimal128`	7,050,299	107.289

Division

Intel Compiler:

Type	Runtime (us)	Ratio to `double`
`float`	100,799	0.971
`double`	103,796	1.000
`decimal32_t`	2,134,312	20.563
`decimal64_t`	5,399,276	52.018
`decimal128_t`	10,012,578	96.464
`decimal_fast32_t`	1,558,774	15.018
`decimal_fast64_t`	1,597,873	15.394
`decimal_fast128_t`	8,105,004	78.086
Intel `Decimal32`	1,561,213	15.041
Intel `Decimal64`	3,115,862	30.019
Intel `Decimal128`	7,474,712	72.013

GCC:

Type	Runtime (us)	Ratio to `double`
`float`	60,277	0.747
`double`	80,676	1.000
`decimal32_t`	2,396,732	29.708
`decimal64_t`	4,021,720	49.850
`decimal128_t`	10,677,625	132.352
`decimal_fast32_t`	1,083,011	13.424
`decimal_fast64_t`	1,851,520	22.950
`decimal_fast128_t`	8,121,160	100.664
GCC `_Decimal32`	5,082,812	63.002
GCC `_Decimal64`	3,005,153	37.250
GCC `_Decimal128`	10,257,437	130.490

`from_chars`

General Format

Type	Runtime (us)	Ratio to `double`
`float`	2,437,788	0.917
`double`	2,657,378	1.000
`decimal32_t`	3,131,251	1.178
`decimal64_t`	4,291,891	1.615
`decimal128_t`	9,911,474	3.730
`decimal_fast32_t`	4,737,095	1.783
`decimal_fast64_t`	4,404,334	1.657
`decimal_fast128_t`	10,414,943	3.919

Scientific Format

Type	Runtime (us)	Ratio to `double`
`float`	2,506,008	0.954
`double`	2,625,702	1.000
`decimal32_t`	3,008,653	1.146
`decimal64_t`	4,180,192	1.592
`decimal128_t`	9,712,229	3.699
`decimal_fast32_t`	4,142,588	1.578
`decimal_fast64_t`	4,118,461	1.569
`decimal_fast128_t`	8,772,097	3.341

`to_chars`

General Format Shortest Precision

Type	Runtime (us)	Ratio to `double`
`float`	2,920,036	0.850
`double`	3,436,919	1.000
`decimal32_t`	4,136,631	1.204
`decimal64_t`	4,318,996	1.257
`decimal128_t`	14,624,180	4.255
`decimal_fast32_t`	4,752,219	1.383
`decimal_fast64_t`	4,382,014	1.275
`decimal_fast128_t`	17,350,588	5.048

General Format 6 digits Precision

Type	Runtime (us)	Ratio to `double`
`float`	5,541,073	0.969
`double`	5,716,626	1.000
`decimal32_t`	3,527,433	0.617
`decimal64_t`	4,125,772	0.722
`decimal128_t`	6,967,211	1.219
`decimal_fast32_t`	3,654,219	0.639
`decimal_fast64_t`	3,386,125	0.592
`decimal_fast128_t`	6,018,439	1.053

Scientific Format Shortest Precision

Type	Runtime (us)	Ratio to `double`
`float`	2,841,569	0.827
`double`	3,437,387	1.000
`decimal32_t`	2,564,053	0.750
`decimal64_t`	2,856,944	0.831
`decimal128_t`	12,147,039	3.534
`decimal_fast32_t`	2,878,507	0.837
`decimal_fast64_t`	2,933,273	0.853
`decimal_fast128_t`	15,010,374	4.367

Scientific Format 6 digits Precision

Type	Runtime (us)	Ratio to `double`
`float`	4,896,523	0.958
`double`	5,112,924	1.000
`decimal32_t`	2,542,237	0.497
`decimal64_t`	3,119,552	0.610
`decimal128_t`	4,811,741	0.941
`decimal_fast32_t`	2,890,023	0.565
`decimal_fast64_t`	2,956,466	0.578
`decimal_fast128_t`	5,476,431	1.071

x64 Windows

Run using an Intel i9-11900k chipset running Windows 11 and Visual Studio 17.14.10

Comparisons

Type	Runtime (us)	Ratio to `double`
`float`	191,653	1.028
`double`	186,424	1.000
`decimal32_t`	2,391,863	12.830
`decimal64_t`	2,491,239	13.363
`decimal128_t`	16,643,031	89.275
`decimal_fast32_t`	872,997	4.682
`decimal_fast64_t`	793,997	4.259
`decimal_fast128_t`	801,708	4.300

Addition

Type	Runtime (us)	Ratio to `double`
`float`	76,777	0.961
`double`	79,897	1.000
`decimal32_t`	2,902,356	36.326
`decimal64_t`	3,569,820	44.680
`decimal128_t`	12,075,529	151.139
`decimal_fast32_t`	1,940,333	24.285
`decimal_fast64_t`	3,064,073	38.350
`decimal_fast128_t`	3,109,101	38.914

Subtraction

Type	Runtime (us)	Ratio to `double`
`float`	336,960	1.042
`double`	323,282	1.000
`decimal32_t`	3,040,167	9.404
`decimal64_t`	3,617,843	11.191
`decimal128_t`	12,325,962	38.128
`decimal_fast32_t`	2,313,234	7.155
`decimal_fast64_t`	2,935,476	9.080
`decimal_fast128_t`	2,963,570	9.167

Multiplication

Type	Runtime (us)	Ratio to `double`
`float`	78,634	1.000
`double`	78,649	1.000
`decimal32_t`	2,636,784	33.526
`decimal64_t`	3,861,139	49.093
`decimal128_t`	11,349,378	144.304
`decimal_fast32_t`	2,688,661	34.186
`decimal_fast64_t`	3,504,172	44.554
`decimal_fast128_t`	9,236,110	117.434

Division

Type	Runtime (us)	Ratio to `double`
`float`	83,566	0.936
`double`	89,317	1.000
`decimal32_t`	3,048,254	34.128
`decimal64_t`	3,282,819	36.755
`decimal128_t`	16,648,799	186.401
`decimal_fast32_t`	2,059,743	23.061
`decimal_fast64_t`	5,105,018	57.156
`decimal_fast128_t`	11,587,763	129,737

`from_chars`

General Format

Type	Runtime (us)	Ratio to `double`
`float`	7,892,780	0.457
`double`	17,282,516	1.000
`decimal32_t`	3,544,166	0.205
`decimal64_t`	5,095,337	0.295
`decimal128_t`	18,972,286	1.098
`decimal_fast32_t`	5,182,044	0.300
`decimal_fast64_t`	6,344,823	0.367
`decimal_fast128_t`	34,476,545	1.995

Scientific Format

Type	Runtime (us)	Ratio to `double`
`float`	7,839,980	0.454
`double`	17,282,516	1.000
`decimal32_t`	3,393,317	0.196
`decimal64_t`	4,924,720	0.285
`decimal128_t`	29,240,187	1.692
`decimal_fast32_t`	5,092,334	0.295
`decimal_fast64_t`	6,341,230	0.367
`decimal_fast128_t`	34,519,610	1.997

`to_chars`

General Format Shortest Precision

Type	Runtime (us)	Ratio to `double`
`float`	3,181,029	0.826
`double`	3,852,857	1.000
`decimal32_t`	5,242,934	1.361
`decimal64_t`	5,586,541	1.450
`decimal128_t`	13,955,214	3.622
`decimal_fast32_t`	6,053,804	1.571
`decimal_fast64_t`	7,957,278	2.065
`decimal_fast128_t`	20,202,107	5.243

General Format 6 digits Precision

Type	Runtime (us)	Ratio to `double`
`float`	6,111,231	0.949
`double`	6,433,885	1.000
`decimal32_t`	4,605,311	0.716
`decimal64_t`	4,742,497	0.737
`decimal128_t`	12,372,901	1.923
`decimal_fast32_t`	4,716,827	0.733
`decimal_fast64_t`	4,861,975	0.756
`decimal_fast128_t`	10,779,778	1.675

Scientific Format Shortest Precision

Type	Runtime (us)	Ratio to `double`
`float`	3,107,509	0.773
`double`	4,020,767	1.000
`decimal32_t`	3,428,517	0.853
`decimal64_t`	4,095,802	1.019
`decimal128_t`	11,577,791	2.879
`decimal_fast32_t`	3,375,975	0.840
`decimal_fast64_t`	4,427,563	1.101
`decimal_fast128_t`	13,581,654	3.378

Scientific Format 6 digits Precision

Type	Runtime (us)	Ratio to `double`
`float`	4,938,623	0.930
`double`	5,309,818	1.000
`decimal32_t`	3,435,843	0.647
`decimal64_t`	3,682,980	0.694
`decimal128_t`	9,223,227	1.737
`decimal_fast32_t`	3,379,702	0.637
`decimal_fast64_t`	3,892,990	0.733
`decimal_fast128_t`	10,158,657	1.913

ARM64 macOS

Run using a Macbook pro with M4 Max chipset running macOS Sequoia 15.5 and homebrew Clang 20.1.8

Comparisons

Type	Runtime (us)	Ratio to `double`
`float`	64,639	1.606
`double`	40,255	1.000
`decimal32_t`	957,179	23.778
`decimal64_t`	897,409	22.293
`decimal128_t`	2,131,391	52.947
`decimal_fast32_t`	380,892	9.462
`decimal_fast64_t`	481,455	11.960
`decimal_fast128_t`	465,461	11.563

Addition

Type	Runtime (us)	Ratio to `double`
`float`	11,853	0.964
`double`	12,295	1.000
`decimal32_t`	1,338,796	108.889
`decimal64_t`	1,231,462	100.160
`decimal128_t`	2,262,808	184.043
`decimal_fast32_t`	608,660	49.505
`decimal_fast64_t`	847,512	68.931
`decimal_fast128_t`	1,030,662	83.827

Subtraction

Type	Runtime (us)	Ratio to `double`
`float`	11,939	0.951
`double`	12,551	1.000
`decimal32_t`	1,296,430	103.293
`decimal64_t`	1,180,456	94.053
`decimal128_t`	2,078,008	165.565
`decimal_fast32_t`	817,989	65.173
`decimal_fast64_t`	823,569	65.618
`decimal_fast128_t`	993,447	79.153

Multiplication

Type	Runtime (us)	Ratio to `double`
`float`	12,186	0.944
`double`	12,914	1.000
`decimal32_t`	1,441,141	111.595
`decimal64_t`	2,117,061	163.935
`decimal128_t`	5,376,470	416.329
`decimal_fast32_t`	923,346	71.500
`decimal_fast64_t`	1,766,419	136.783
`decimal_fast128_t`	5,463,675	423.082

Division

Type	Runtime (us)	Ratio to `double`
`float`	12,576	0.722
`double`	17,145	1.000
`decimal32_t`	1,732,611	101.056
`decimal64_t`	3,558,094	207.529
`decimal128_t`	8,985,521	524.090
`decimal_fast32_t`	1,075,184	62.711
`decimal_fast64_t`	2,027,533	118.258
`decimal_fast128_t`	7,583,016	442.287

`from_chars`

General Format

Type	Runtime (us)	Ratio to `double`
`float`	1,882,825	0.990
`double`	1,901,380	1.000
`decimal32_t`	3,427,654	1.803
`decimal64_t`	5,364,564	2.821
`decimal128_t`	11,752,375	6.181
`decimal_fast32_t`	4,339,550	2.282
`decimal_fast64_t`	6,647,959	3.496
`decimal_fast128_t`	14,010,588	7.369

Scientific Format

Type	Runtime (us)	Ratio to `double`
`float`	1,939,033	1.010
`double`	1,919,671	1.000
`decimal32_t`	3,411,016	1.777
`decimal64_t`	5,278,214	2.750
`decimal128_t`	11,479,704	5.980
`decimal_fast32_t`	4,299,497	2.240
`decimal_fast64_t`	6,287,638	3.275
`decimal_fast128_t`	9,856,122	5.134

`to_chars`

General Format Shortest Precision

Type	Runtime (us)	Ratio to `double`
`float`	2,223,891	0.882
`double`	2,520,203	1.000
`decimal32_t`	2,983,523	1.184
`decimal64_t`	3,348,702	1.329
`decimal128_t`	8,899,289	3.531
`decimal_fast32_t`	3,383,567	1.343
`decimal_fast64_t`	3,436,470	1.364
`decimal_fast128_t`	12,509,443	4.964

General Format 6 digits Precision

Type	Runtime (us)	Ratio to `double`
`float`	4,664,538	0.948
`double`	4,915,699	1.000
`decimal32_t`	2,570,339	0.523
`decimal64_t`	3,309,343	0.673
`decimal128_t`	5,962,030	1.212
`decimal_fast32_t`	2,213,792	0.450
`decimal_fast64_t`	3,067,584	0.624
`decimal_fast128_t`	6,006,157	1.222

Scientific Format Shortest Precision

Type	Runtime (us)	Ratio to `double`
`float`	2,119,538	0.848
`double`	2,500,900	1.000
`decimal32_t`	1,757,416	0.703
`decimal64_t`	2,187,911	0.875
`decimal128_t`	6,976,380	2.790
`decimal_fast32_t`	1,739,069	0.695
`decimal_fast64_t`	2,060,848	0.824
`decimal_fast128_t`	12,509,443	5.002

Scientific Format 6 digits Precision

Type	Runtime (us)	Ratio to `double`
`float`	4,157,977	0.933
`double`	4,457,878	1.000
`decimal32_t`	1,764,018	0.395
`decimal64_t`	2,625,621	0.589
`decimal128_t`	4,060,487	0.911
`decimal_fast32_t`	1,728,473	0.388
`decimal_fast64_t`	2,734,955	0.614
`decimal_fast128_t`	5,300,774	1.189

Edit this Page